Particularly in the design phase, it can be difficult to make well-founded statements about the expected operating costs.
The CLI tool Infracost starts right here and enables the automated estimation of cloud infrastructure costs based on existing Terraform resources.
As cloud infrastructure is generally not managed manually, but via Infrastructure as Code (IaC) tools such as Terraform should be defined, created and managed, the Terraform resources also reflect the actual cloud infrastructure very well. In order to estimate the costs involved, Infracost compares all Terraform resources used with the Infracost Cloud Pricing API which is updated once a week with the official prices of the cloud providers AWS, Azure and Google.
The advantage of this is that the cloud infrastructure does not even have to actually exist. The definition alone using Terraform is sufficient to obtain a cost estimate, i.e. even before the cloud infrastructure is created using "terraform apply" is rolled out. This allows cost risks to be identified at an early stage, budgets to be planned realistically and decisions to be made based on data. Infracost is therefore ideal for DevOps teams and cloud architects who want to avoid surprises.
How to use Infracost effectively in practice
Use case 1 - Knowing what your cloud costs before the rollout
This use case outlines the situation described above, where Terraform resources are defined but not yet available using "terraform apply" and the cloud infrastructure therefore does not yet exist at the planned cloud provider. In this case, the command "infracost breakdown --path .", which is to be executed in the file directory of the Terraform resources. Infracost produces the following output, for example, which in this example refers to Azure resources:
INFO Autodetected 1 Terraform project across 1 root module
INFO Found Terraform project main at directory . using Terraform var files terraform.tfvars
Project: main
Name Monthly Qty Unit Monthly Cost
module.azurerm_kubernetes.azurerm_kubernetes_cluster.kubernetes_cluster
└─ default_node_pool
├─ Instance usage (Linux, pay as you go, Standard_B2ms) 730 hours $70.08
└─ os_disk
└─ Storage (P10, LRS) 1 months $21.68
module.azurerm_container_registry.azurerm_container_registry.container_registry
├─ Registry usage (Standard) 30 days $20.00
├─ Storage (over 100GB) Monthly cost depends on usage: $0.10 per GB
└─ Build vCPU Monthly cost depends on usage: $0.0001 per seconds
module.azurerm_postgresql_database.azurerm_postgresql_flexible_server.database_server
├─ Compute (B_Standard_B1ms) 730 hours $14.53
├─ Storage 32 GB $4.38
└─ Additional backup storage Monthly cost depends on usage: $0.10 per GB
module.azurerm_kubernetes.azurerm_public_ip.public_ip
└─ IP address (static, regional) 730 hours $3.65
module.azurerm_postgresql_database.azurerm_private_dns_zone.dns_zone
└─ Hosted zone 1 months $0.50
module.azurerm_storage.azurerm_storage_account.storage_account
├─ Capacity Monthly cost depends on usage: $0.0196 per GB
├─ Write operations Monthly cost depends on usage: $0.054 per 10k operations
├─ List and create container operations Monthly cost depends on usage: $0.054 per 10k operations
├─ Read operations Monthly cost depends on usage: $0.0043 per 10k operations
├─ All other operations Monthly cost depends on usage: $0.0043 per 10k operations
└─ Blob index Monthly cost depends on usage: $0.039 per 10k tags
OVERALL TOTAL $134.82
*Usage costs can be estimated by updating Infracost Cloud settings, see docs for other options.
──────────────────────────────────
13 cloud resources were detected:
∙ 6 were estimated
∙ 7 were free
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Project ┃ Baseline cost ┃ Usage cost* ┃ Total cost ┃
┣━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╋━━━━━━━━━━━━━━━╋━━━━━━━━━━━━━╋━━━━━━━━━━━━┫
┃ main ┃ $135 ┃ - ┃ $135 ┃
┗━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┻━━━━━━━━━━━━━━━┻━━━━━━━━━━━━━┻━━━━━━━━━━━━┛
In this example, Infracost estimates the initial total costs of the planned cloud infrastructure per calendar month at $135. By default, however, only the baseline costs are taken into account, i.e. the fixed costs that are incurred regardless of the actual duration or frequency of use of the cloud infrastructure. If you do not specify any concrete quantities, the usage costs are not included in the estimate. This can lead to very inaccurate cost estimates, especially for data-driven projects, as the costs scale greatly with the amount of data: The more data is received, processed and stored via the cloud infrastructure at runtime, the higher its costs. Therefore, if a quantity structure that is as concrete as possible is already known, this should be used via the infracost-usage.yml can be specified (see Usage costs).
If you want to redirect each cost estimate to a JSON file and version it via Git, for example, you can use the "infracost breakdown --path . --format json --out-file <costs-file>" can be used, which is particularly interesting in Use Case 3.
Use Case 2 - How expensive is my current cloud infrastructure
This use case outlines the case where Terraform resources have already been applied, but no changes have been made to them. Here too, the command "infracost breakdown --path .", which should lead to the same output as in Use Case 1, provided the Terraform resources have not been changed in the meantime.
Important: As soon as actual costs are incurred by the cloud provider, these costs should be regularly compared with the costs estimated by Infracost in order to validate them. If the actual costs are higher than the estimated costs, either cloud services are being used that have not been defined using Terraform and are therefore unknown to Infracost, or the usage costs have not been taken into account at all or only in part. In general, it should be noted that in all cases it is only a cost estimate, which in the best case differs only minimally from the actual costs.
Use Case 3 - Changing costs: How do changes affect my costs?
This use case outlines the case where changes are made to Terraform resources that have already been applied, which affect the costs of the cloud infrastructure. In the following example, a Kubernetes cluster node size was increased from "Standard_B2ms" on "Standard_B4ms". The command "infracost breakdown --path ." for use:
INFO Autodetected 1 Terraform project across 1 root module
INFO Found Terraform project main at directory . using Terraform var files terraform.tfvars
Project: main
Name Monthly Qty Unit Monthly Cost
module.azurerm_kubernetes.azurerm_kubernetes_cluster.kubernetes_cluster
└─ default_node_pool
├─ Instance usage (Linux, pay as you go, Standard_B4ms) 730 hours $140.16
└─ os_disk
└─ Storage (P10, LRS) 1 months $21.68
module.azurerm_container_registry.azurerm_container_registry.container_registry
├─ Registry usage (Standard) 30 days $20.00
├─ Storage (over 100GB) Monthly cost depends on usage: $0.10 per GB
└─ Build vCPU Monthly cost depends on usage: $0.0001 per seconds
module.azurerm_postgresql_database.azurerm_postgresql_flexible_server.database_server
├─ Compute (B_Standard_B1ms) 730 hours $14.53
├─ Storage 32 GB $4.38
└─ Additional backup storage Monthly cost depends on usage: $0.10 per GB
module.azurerm_kubernetes.azurerm_public_ip.public_ip
└─ IP address (static, regional) 730 hours $3.65
module.azurerm_postgresql_database.azurerm_private_dns_zone.dns_zone
└─ Hosted zone 1 months $0.50
module.azurerm_storage.azurerm_storage_account.storage_account
├─ Capacity Monthly cost depends on usage: $0.0196 per GB
├─ Write operations Monthly cost depends on usage: $0.054 per 10k operations
├─ List and create container operations Monthly cost depends on usage: $0.054 per 10k operations
├─ Read operations Monthly cost depends on usage: $0.0043 per 10k operations
├─ All other operations Monthly cost depends on usage: $0.0043 per 10k operations
└─ Blob index Monthly cost depends on usage: $0.039 per 10k tags
OVERALL TOTAL $204.90
*Usage costs can be estimated by updating Infracost Cloud settings, see docs for other options.
──────────────────────────────────
13 cloud resources were detected:
∙ 6 were estimated
∙ 7 were free
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Project ┃ Baseline cost ┃ Usage cost* ┃ Total cost ┃
┣━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╋━━━━━━━━━━━━━━━╋━━━━━━━━━━━━━╋━━━━━━━━━━━━┫
┃ main ┃ $205 ┃ - ┃ $205 ┃
┗━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┻━━━━━━━━━━━━━━━┻━━━━━━━━━━━━━┻━━━━━━━━━━━━┛
Infracost now estimates the costs of the changed cloud infrastructure per calendar month at $205, which represents an increase of $70 compared to Use Case 2.
To avoid having to calculate the cost difference yourself, you can alternatively use "infracost diff --path <new-costs-file> --compare-to <old-costs-file>" which leads to the following output:
Key: * usage cost, ~ changed, + added, - removed
──────────────────────────────────
Project: main
~ module.azurerm_kubernetes.azurerm_kubernetes_cluster.kubernetes_cluster
+$70 ($92 → $162)
~ default_node_pool
~ Instance usage (Linux, pay as you go, Standard_B2ms → Standard_B4ms)
+$70 ($70 → $140)
Monthly cost change for Kunde_doubleSlash/update-manager/additional/dsum-infrastructure/terraform/environments/dev
Amount: +$70 ($135 → $205)
Percent: +52%
──────────────────────────────────
Key: * usage cost, ~ changed, + added, - removed
*Usage costs can be estimated by updating Infracost Cloud settings, see docs for other options.
13 cloud resources were detected:
∙ 6 were estimated
∙ 7 were free
Infracost estimate: Monthly estimate increased by $70 ↑
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┓
┃ Changed project ┃ Baseline cost ┃ Usage cost* ┃ Total change ┃
┣━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╋━━━━━━━━━━━━━━━╋━━━━━━━━━━━━━╋━━━━━━━━━━━━━━┫
┃ Kunde_doubleSlash/update-manage...ure/terraform/environments/dev ┃ +$70 ┃ - ┃ +$70 (+52%) ┃
┗━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┻━━━━━━━━━━━━━━━┻━━━━━━━━━━━━━┻━━━━━━━━━━━━━━┛
This approach is particularly useful for the use of Infracost in an automated CI/CD pipeline, whereby the older of the two JSON files can be versioned using Git, for example, and compared with a newly created JSON file. For the next execution of the pipeline, the versioned JSON file must then be overwritten with the newly created JSON file so that it contains the cost estimate of the Terraform resources actually used.
Not everyone can know everything
Infracost currently supports around 1,100 Terraform resources. If you use services that are not (yet) supported, the estimate may be incomplete.
Avoid unpleasant surprises with Infracost
Infracost offers enormous added value when it comes to proactively planning cloud infrastructure costs in conjunction with Terraform. Who Infrastructure as Code (IaC) should also think "Costs as Code". This is exactly what Infracost makes possible.
For two reasons, however, the following should always apply: Trust is good, control is better!
Reason 1: A cost estimate is just a cost estimate and does not necessarily have to correspond to the actual costs.
Reason 2: A cost estimate is only as good as the information that has been taken into account. A missing or inadequate quantity structure can greatly distort the cost estimate due to incorrect usage costs.
